romanoviolet / udacity-localization

An implementation of Markov Localization Algorithm based on lectures from the Udacity's Self Driving Care Nano Degree (SDCND) Program

(Markov) Localization

About

This project implements Markov Localization algorithm as required by the Udacity's Self Driving Car Nano-Degree program.

The input to the algorithm are the position of a set of landmarks observed by the radar (or the lidar) mounted on the vehicle as it moves along a path. In addition, the vehicle is provided with a sparse map containing the position of a set of features. As the vehicle moves, the (noisy) sensors sense the location of objects in the vicinity, and the localization algorithm uses the data provided by the sensors, along with the map in order to precisely localize the vehicle. The localization is probabilistic in nature, and is based on particle filters. A rough estimate of the initial position of the vehicle is provided by a sensor (e.g., GPS). Thereafter, the localization is based solely on radar (and/or lidar) sensor data, along with the information contained in the map.

How it Looks

This project involves the Term 2 Simulator which can be downloaded here. Select "Kidnapped Vehicle" option.

A sample frame from grabbed as the vehicle localizes itself inside a simulator is shown The accuracy of the localizer is indicated by the blue circle, (almost) overlaid over the simulated car. Udacity also imposed requirements on the performance (i.e., speed) of the implementation, which is enforced by the simulator. The simulator expects that one complete simulation required less than 100 seconds.

Prerequisites

Ad-verbatim from Udacity's Instructions:

uWebSocketIO Starter Guide

All of the projects in Term 2 and some in Term 3 involve using an open source package called uWebSocketIO. This package facilitates the same connection between the simulator and code that was used in the Term 1 Behavioral Cloning Project, but now with C++. The package does this by setting up a web socket server connection from the C++ program to the simulator, which acts as the host. In the project repository there are two scripts for installing uWebSocketIO - one for Linux and the other for macOS.

Note: Only uWebSocketIO branch e94b6e1, which the scripts reference, is compatible with the package installation. Linux Installation:

From the project repository directory run the script: install-ubuntu.sh

Structure of the Project

The project is structured as follows:

• Folder "src": Contains the core logic required to build the localizer based on particle filters
• Folder "Results": Contains a sample video and a screenshot when testing the localizer using the simulator supplied by Udacity.
• clean.sh: A shell script to delete an old build (if it exists)
• build.sh: A shell script to build the Localizer project.
• install-ubuntu.sh: A script required for installing dependencies for running the simulator. See the section above on "Prerequisites"
• run.sh: A shell script to run the binary built using build.sh, and connects to the Udacity's Term 2 simulator. See the section above "How it Looks".

Running the Localizer

Execute the run.sh which makes the Localizer application wait on the Udacity's Term 2 simulator. When running the simulator, choose the option "Kidnapped Vehicle".

Pending Improvements

• Improvements to the accuracy of the Localizer Filter are possible, and are welcome.

Credits

• Udacity: Lecturers, and mentors;
• Internet: for examples and samples.

Disclaimer

Some of the ideas are borrowed and adapted from other people's work.